SINGLE VARIABLE MODEL

PREDICTIVE CONTROL
TEAM 6

INTRODUCTION

Most modifications to single-loop feedback control presented in this model have used
additional measurements to improve control performance.
This is an alternative to the proportional-integral-derivative (PID) feedback algorithm.

PID

IMC

THE MODEL PREDICTIVE CONTROL

STRUCTURE

D(s)

Gd(s)

SP(s)
+

Tp(s)

CV(s)

+
Gep(s)

Gp(s)

+
-

MV(s)

Gm(s)

Em(s)

THE IMC CONTROLLER (Internal Model

Control)
The IMC approach segregates and eliminates the aspects of the model transfer function
that make calculation of a realizable inverse impossible. The first step is to factor the model
into the product of the two factors:
+ () ()
=

()

the
noninvertible
part
has
an
the invertible part has an inverse that is stable

inverse

that

is

unstable

The IMC controller eliminates all elements in the process model Gm(s) that lead to an
unrealizable controller by taking the inverse of only the invertible factor to give
()
=

Structure Controller IMC

Controller

Process

Process
Model

Process
Model

For a set point change

 The IMC controller is based on the general predictive control structure. The controller
design method adheres to criteria that ensure zero offset for sleeplike disturbances, and it
employs a factorization approach to obtain a realizable approximate inverse that gives
good feedback control performance. An adjustable filter (tuning parameter) was
introduced to enable the engineer to moderate the feedback action to maintain good
performance of the controller and manipulated variables in the presence of
measurement noise and model error.

The Smith Predictor

The control designed by O. Smith preceded much of the general analysis of predictive
systems; in fact; it predated the application of digital computers to process control, so that
widespread implementation of Smiths results were delayed until real-time digital control
computers became commercially available.

Smiths approach, shown in

the figure, relies on the
general predictive structure
in which the controller is
calculated by the elements
in the dashed box; these
elements perform the
function of the predictive
control algorithm Gcp(s) in
the last figure

 In conclusion, the Smith predictor conforms to the general principles of the predictive
control structure. It employs a unique method for calculating an approximate model
inverse: by controlling a model consisting of the invertible part of the model. This structure
can achieve zero steady state offset for steplike disturbances by conforming to easily
achieved criteria.
Again, the Smith predictor system is simple to implement in digital control and generally
yields good control performance
The tuning of the PI controller must be appropriate for the predictive structure and can be
adjust to make the Smith Predictor more or less aggressive to provide the desired
controlled and manipulated variable performance for the expected range of model
mismatch

Implementation Guidelines
It is important to remember that predictive controllers employ the same feedback principles as
classical structures and involve basically the same task to design, implement, and operate.
The engineering tasks include:
-Selecting the feedback measurement.
-Selecting the manipulated variable.
-Determine an appropriate model structure with parameters.
-Selecting an algorithm.
-Establishing the tuning contents.
-Decide the status of the controller (automatic or manual).
-Enter the set point value.

Incorrect(Windup occurs)
1

lim () = lim

1
0
0
1 ( )

Correct(Windup prevented)
1
lim () = ( )
0

THANK YOU
:D